Cloud-based frontier AI models operated by OpenAI, Anthropic, xAI, and Google experienced a rare, overlapping set of service outages on Thursday morning. Anthropic reported elevated errors across several models including Claude Opus 5 starting at 9:23 a.m. Eastern before deploying a fix by midday. Simultaneously, OpenAI flagged degraded performance across ChatGPT and Codex beginning at 10:43 a.m., while user report spikes indicated concurrent outages for xAI’s Grok and Google’s Gemini API.
Major underlying cloud infrastructure providers including Amazon Web Services, Microsoft Azure, and Cloudflare did not report widespread system failures during the timeframe, making the simultaneous frontier model downtime unusual given their historically high uptime metrics.
All four companies resolved the issues or saw service levels return to normal by early Thursday afternoon, though the causes behind the concurrent disruptions remain unconfirmed.
Why it matters
Highlights reliability risks for enterprise operators relying heavily on single-cloud or single-model API setups for critical production workloads.
Emphasizes the necessity of multi-model failover architectures to maintain service uptime during unexpected vendor outages.
Source: arstechnica.com
